主讲人：Prof. Ligang He
讲座方式与地点：网络视频，腾讯会议ID：189 678 673
Title: Federated Learning – Part 1 Workflow Optimization
Abstract: The first talk of this talk series will cover two parts: 1) the background of Federated Learning (FL) and 2) our own work in FL. In the first part, the training workflow of FL will be introduced. Two main types of FL protocols, i.e., synchronous and asynchronous protocols, and their pros and cons will be presented. The relation between FL and other technologies, such as other machine learning techniques and distributed computing, will be discussed. In the second part, our own work in FL will be presented. In particular, we develop a new semi-asynchronous training workflow for FL called SAFA. SAFA enables flexible device participation. The FL server can synchronize with tolerance, while clients have the chance to keep partially-trained local modes to reduce the “waste” of local training. Some evaluation results of SAFA will also be presented.
Dr. Ligang He现为英国华威大学（University of Warwick ）计算机系终身教授，在学术界具有较高的声誉，在国际著名期刊（例如IEEE Transactions on Parallel and Distributed Systems, IEEE Transactions on Computers，EEE Transactions on Network Science and Engineering、IEEE Transactions on Cloud Computing. IEEE Transactions on Services Computing, ACM Transactions on Computer Systems等）和重要会议上(例如VLDB，IPDPS, SC, HPCA, ICPP, ICSOC等)发表上述领域的论文140余篇。得到来自英国EPRSC、Leverhulme，欧盟及工业界的多项科研资助。研究方向为大数据分析、并行分布式计算和云计算。